Summary: Robust Planning with (L)RTDP
Olivier Buffet and Douglas Aberdeen
National ICT Australia &
The Australian National University
{olivier.buffet,douglas.aberdeen}@nicta.com.au
Abstract
Stochastic Shortest Path problems (SSPs), a sub­
class of Markov Decision Problems (MDPs), can
be efficiently dealt with using Real­Time Dynamic
Programming (RTDP). Yet, MDP models are often
uncertain (obtained through statistics or guessing).
The usual approach is robust planning: searching
for the best policy under the worst model. This
paper shows how RTDP can be made robust in
the common case where transition probabilities are
known to lie in a given interval.
1 Introduction
In decision­theoretic planning, Markov Decision Problems
[Bertsekas and Tsitsiklis, 1996] are of major interest when
a probabilistic model of the domain is available. A number of

Source: Aberdeen, Douglas - National ICT Australia & Computer Sciences Laboratory, Australian National University

Collections: Computer Technologies and Information Sciences